Skip to content

Observability

Current observability is logging-centric; metrics/tracing are future enhancements.

Logging

  • Console logger configured with information level by default
  • Key events: provider start/end, file processed, chunk counts, errors

Suggested Conventions

Event Level
Start indexing Information
Skip unchanged Information
File processed Information
Unsupported file Warning
Extraction failure Error
Fatal pipeline error Error

Adding Metrics (Future)

Potential integration with OpenTelemetry: - Counter: files_processed - Counter: chunks_indexed - Histogram: embedding_latency_ms - Gauge: providers_enabled

Tracing (Future)

  • Activity per provider run
  • Child spans: list, download, extract, embed, upsert

Log Enrichment

  • Include provider type/name, filename, operation durations

Shipping Logs

  • Container platform aggregation (stdout) -> Loki, ELK, Cloud-native logging

Alert Ideas

Condition Alert
0 files processed Investigate provider config
Error rate spike Extraction / API failure
High embedding latency Upstream model slowness

Next